ABOUT THE PROJECT
The Ultrafast Imaging Senior Capstone aims to improve upon the optical coherence tomography (OCT) system established by previous teams of recent years. The system is based upon the principle of interferometry in which pulses of light are superimposed to create inference. This interference holds information about a given sample that is then processed in real-time using Fast Fourier Transforms (FFT) via Python to produce a 3D image. The goal of this year is to boost the system’s frame rate from 4 fps to 24+ fps as well as increase the image resolution from 100 volumetric pixels to 200+. The main steps towards achieving this goal involve resolving the current hardware limitations in the system, implementing a new PC with high-end components, including a discrete GPU, and reworking the code to utilize GPU-acceleration.
LIMITATIONS
Current System Bottlenecks
SCANNER
2 kHz x 20Â Hz
The system scans bilaterally at a 100:1 ratio to create a 100 volumetric pixel image. Currently, both scanners in the system oscillate a max frequency of 2 kHz. The faster scanner quickly oscillates across the sample taking 100 points while the slower scanner moves 1 point orthogonally. Due to the max frequency of the instruments, the frame rate and resolution of the image are inhibited.
To alleviate this, we plan to install a new scanner, rated at 500 kHz, to replace the faster scanner. While this stands to greatly increase scanning speed, the field of view (FOV) drastically decreases. This will lead us into another problem to solve.
OSCILLOSCOPE
Bandwidth/Sampling Rate v Data Transfer Speed
There are 3 oscilloscopes available to us in the lab. The options are the Tektronix DSA72004C rated at 20 GHz, 100 GS/s, a Tektronix TDS6804B rated at 8 GHz, 20 GS/s, and a Tektronix MSO54 rated at 1 GHz, 6.25 GS/s. The limitations are seeded in the link speeds of each instrument. The DSA72004C is damaged and the Ethernet can only transfer data at 100 Mb/s. The TDS6804B has a max transfer speed of 1 Gb/s through Ethernet. Finally, the MSO54 has a max transfer rate of 5 Gb/s through a USB 3.0 Type-A port.
​
Our plan is to calculate and test the drawbacks of each oscilloscope to determine the best choice for this project.
PROCESSING
4 fps
The system renders 3D images at 4 fps. This is primarily due to the Intel i5-6500 central processing unit performing the FFTs. This CPU is three generations behind and is not well suited to handle these calculations continuously with large amounts of data.
With a new PC comprised of an AMD Ryzen 9 3900X, an NVIDIA RTX-2080 Super GPU, and ample amounts of high-speed RAM, the overall system will see significant improvement. While the newer CPU gives a substantial boost out of the box, the goal will not be fully realized until the GPU is properly integrated. A GPU-compatible code must be created, either in CUDA or PyCUDA, to utilize the GPU's processing power.
TEAM
TREVOR WONG
Software Lead
An undergraduate computer engineering student at Loyola Marymount University and part of the team developing the Ultrafast Imaging project. While both members are tasked with becoming experts with the system, Trevor will be taking the lead on GPU-accelerating.
ALEX ESCLAMADO
Hardware Lead
An undergraduate computer engineering student at Loyola Marymount University and part of the team developing the Ultrafast Imaging project. While both members are tasked with becoming experts with the system, Alex will be taking the lead on hardware alterations.